Construction and validation of a prognostic nomogram in metastatic breast cancer patients of childbearing age: A study based on the SEER database and a Chinese cohort

Abstract

Introduction: Cancer in patients of childbearing age continues to become increasingly common. The purpose of this study was to explore the impact of metastatic breast cancer (MBC) on overall survival (OS) and cancer-specifific survival (CSS) in patients of childbearing age and to construct prognostic nomograms to predict OS and CSS.

Methods: Data from MBC patients of childbearing age were obtained from the Surveillance, Epidemiology, and End Results (SEER) database between 2010 and 2015, and the patients were randomly assigned into the training and validation cohorts. Univariate and multivariate Cox analyses were used to search for independent prognostic factors impacting OS and CSS, and these data were used to construct nomograms. The concordance index (C-index), area under the curve (AUC), and calibration curves were used to determine the predictive accuracy and discriminative ability of the nomograms. Additional data were obtained from patients at the Yunnan Cancer Hospital to further verify the accuracy of the nomograms.

Results: A total of 1,700 MBC patients of childbearing age were identifified from the SEER database, and an additional 92 eligible patients were enrolled at the Yunnan Cancer Hospital. Multivariate Cox analyses identifified 10 prognostic factors for OS and CSS that were used to construct the nomograms. The calibration curve for the probabilities of OS and CSS showed good agreement between nomogram prediction and clinical observations. The C-index of the nomogram for OS was 0.735 (95% CI = 0.725-0.744); the AUC at 3 years was 0.806 and 0.794 at 5 years.The nomogram predicted that the C-index of the CSS was 0.740 (95% CI = 0.730- 0.750); the AUC at 3 years was 0.811 and 0.789 at 5 years. The same results were observed in the validation cohort. Kaplan- Meier curves comparing the low-,medium-, and high-risk groups showed strong prediction results for the prognostic nomogram.

Conclusion: We identifified several independent prognostic factors and constructed nomograms to predict the OS and CSS for MBC patients of childbearing age.These prognostic models should be considered in clinical practice to individualize treatments for this group of patients.

Keywords: SEER; childbearing age; females; metastatic breast cancer (mbc); nomogram; prognosis.

Figure 1

A flow diagram showing the screening process for the analysis of patients in the SEER and Yunnan cohorts.

Figure 2

Univariate (A) and multivariate (B) Cox proportional hazards regression analysis of OS in the training cohort. Univariate (C) and multivariate (D) Cox proportional hazards regression analysis of CSS in the training cohort. 1: American Indian/AK Native, Asian/Pacific Islander; 2: central portion of breast or nipple.

Figure 3

The prognostic nomograms for OS (A) and CSS (B) in MBC patients of childbearing age in the training cohort. Example of the nomogram. The nomogram can be used to calculate the prediction probability of OS and CSS. The nomogram shows the influence of different prediction variables. The influence of each variable is represented by the horizontal lines, with longer lines indicating a greater impact. The influence of each variable is visualized through multiple points on the corresponding horizontal line. By adding points related to each variable, the expected score size can be read on the response horizontal line at the bottom of the nomogram.

Figure 4

The ROC curve of the nomogram in the training cohort. (A) The AUC for OS at 3 years was 0.806 and at 5 years was 0.794. (B) The AUC for CSS at 3 years was 0.811 and at 5 years was 0.798. (C) The calibration curves for OS of the nomograms. (D) The calibration curves for CSS of the nomograms.

Figure 5

Kaplan–Meier analysis of patients in the training cohort. The survival curves were generated from the score calculated by the nomograph for OS (A) and CSS (B). In patients with luminal A BC, the survival curve was generated from the score calculated by the nomograph: OS (C) and CSS (D). In patients with the luminal B subtype, the survival curve was generated from the score calculated by the nomograph: OS (E) and CSS (F). In patients with the HER2-enriched subtype, the survival curve was generated from the score calculated by the nomograph: OS (G) and CSS (H). In patients with the triple-negative subtype, the survival curve was generated from the score calculated by the nomograph: OS (I) and CSS (J).